List your company

Open Question Answering on MATH500 (test)

0.94AccuracyGRPO
0.84640.87070.8950.9193Feb 10, 2026

Evaluation results

MethodLinks
GRPO2026.020.943,962
FlashThink2026.020.9383,050
AdaptThink2026.020.9382,130
ESTARoptimization=RL2026.020.938635
ESTAR-FTmode=fine-tuned2026.020.9342,401
ESTAR-LITEearly_stopping=classifier-based2026.020.9322,019
Length-Penaltyexplicit_length_penalty=true2026.020.9243,190
O1-Pruner2026.020.9162,856
No-Thinkingreasoning=disabled2026.020.851,139